WO2023071874A1

WO2023071874A1 - Roadside assistance working node determining method and apparatus, electronic device, and storage medium

Info

Publication number: WO2023071874A1
Application number: PCT/CN2022/126059
Authority: WO
Inventors: 洪子梦; 施媛媛
Original assignee: 中移(上海)信息通信科技有限公司; 中移智行网络科技有限公司; 中国移动通信集团有限公司
Priority date: 2021-10-29
Filing date: 2022-10-19
Publication date: 2023-05-04
Also published as: CN113705549B; CN113705549A

Abstract

The present invention provides a roadside assistance working node determining method and apparatus, an electronic device, and a storage medium. The roadside assistance working node determining method comprises: obtaining images collected by a plurality of cameras, wherein the plurality of cameras are located at different positions on a rescue vehicle; and inputting the images into a scene classification and identification model, and outputting a roadside assistance working node. Therefore, the accuracy of an output result of a roadside assistance working node is improved.

Description

Road rescue work node determination method, device, electronic equipment, and storage medium

Cross References to Related Applications

This application is based on a Chinese patent application with application number 202111267060.6 and a filing date of October 29, 2021, and claims the priority of this Chinese patent application. The entire content of this Chinese patent application is hereby incorporated by reference.

technical field

The embodiments of the present application relate to the technical field of communications, and in particular to a method and device for determining a road rescue work node, electronic equipment, and a storage medium.

Background technique

With the continuous development of the automobile market, the incidence of road traffic accidents has also increased, and the automobile road rescue industry has emerged as the times require. In the process of continuous optimization and development of the road rescue industry, a series of norms and requirements have emerged for the management of road rescue personnel and standard process control. However, the current road rescue work node is determined by sensors such as door magnets or power take-offs, which leads to a large error in the determination result of the current road rescue work node.

Contents of the invention

Embodiments of the present application provide a method and device for determining a road rescue work node, an electronic device, and a storage medium, which can improve the accuracy of a determination result of a current road rescue work node.

In order to solve the above problems, the application is implemented as follows:

The embodiment of the present application provides a method for determining road rescue work nodes, including:

acquiring images captured by a plurality of cameras, wherein the plurality of cameras are located at different positions on the rescue vehicle;

The image is input into the scene classification and recognition model, and the road rescue work node is output, wherein the scene classification and recognition model is obtained by training a plurality of pre-acquired image sets.

The embodiment of the present application also provides a road rescue work node determination device, including:

An acquisition module, configured to acquire images collected by multiple cameras, wherein the multiple cameras are located at different positions on the rescue vehicle;

The output module is configured to input the image into the scene classification and recognition model, and output the road rescue work node, wherein the scene classification and recognition model is trained through a plurality of pre-acquired image sets.

The embodiment of the present application also provides an electronic device, including: a transceiver, a memory, a processor, and a program stored on the memory and operable on the processor; the processor is used to read the The program implements the steps in the method described in the aforementioned first aspect.

The embodiment of the present application further provides a readable storage medium for storing a program, and when the program is executed by a processor, the steps in the method described in the aforementioned first aspect are implemented.

In the embodiment of the present application, the images collected by multiple cameras are obtained, wherein the multiple cameras are located at different positions on the rescue vehicle; the images are input into the scene classification recognition model, and the road rescue work node is output, wherein, The scene classification recognition model is obtained by training a plurality of pre-acquired image sets. In this way, the scene classification and recognition model is obtained through pre-acquired multiple image sets, and then the scene classification and recognition model is used to identify the images collected by multiple cameras, so as to output the road rescue work node, because the scene classification model is obtained through pre-acquired multiple images The multiple image sets include various images collected by the camera on the rescue vehicle, so that the output result of the scene classification and recognition model is more accurate, which can reduce the error of the output result of the road rescue work node, that is, improve the road rescue work node. The accuracy of the output results.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that need to be used in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

Fig. 1 is a flow chart of a method for determining a road rescue work node provided by an embodiment of the present application;

FIG. 2 is an effect diagram of a method for determining a road rescue work node provided by an embodiment of the present application;

Fig. 3 is a schematic structural diagram of a device for determining a road rescue work node provided by an embodiment of the present application;

FIG. 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of this application.

The terms "first", "second" and the like in the embodiments of the present application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. Furthermore, the terms "comprising" and "having", as well as any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements is not necessarily limited to the expressly listed instead, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus. In addition, the use of "and/or" in this application means at least one of the connected objects, such as A and/or B and/or C, means that A alone, B alone, C alone, and both A and B exist, Both B and C exist, both A and C exist, and there are 7 situations where A, B, and C all exist.

Please refer to Fig. 1, Fig. 1 is a flow chart of a method for determining a road rescue work node provided by the embodiment of the present application, as shown in Fig. 1, including the following steps:

Step 101, acquiring images collected by multiple cameras.

In the embodiment of the present application, the device for determining a road rescue work node acquires images collected by multiple cameras, where the multiple cameras are located at different positions on the rescue vehicle.

Wherein, the device for determining the road rescue working node may be a terminal with corresponding processing functions on the rescue vehicle. The quantity and installation positions of the cameras are not limited here, for example: at least one side of the chassis of the rescue vehicle, the inner wall of the driver, the outer wall of the driver and the body of the rescue vehicle (the body can include a first side, a second side and The third side, the first side and the third side are opposite sides, and one end of the first side and one end of the third side are respectively connected with the cab, and the second side is respectively connected with the other end of the first side and the other end of the third side. One end is connected, that is, the second side and the driver's cab are opposite sides) can be provided with a camera, and the above-mentioned cameras can be used to collect images.

Wherein, the camera can collect images in real time, and the camera can also collect images at regular intervals. In addition, the camera can only collect images when preset conditions are met, and the above preset conditions can include at least one of the following: receiving a collection instruction, the rescue vehicle is located in the target area (the target area includes a rescue workstation station or a highway), Rescue vehicles are activated.

Step 102: Input the image into the scene classification and recognition model, and output the road rescue work node, wherein the scene classification and recognition model is obtained by training a plurality of pre-acquired image sets.

In the embodiment of the present application, the road rescue work node inputs the image into the scene classification and recognition model, and outputs the road rescue work node, wherein the scene classification and recognition model is obtained by training a plurality of pre-acquired image sets.

In the embodiment of the present application, the scene classification and recognition model is obtained through pre-acquired multiple image sets training, and then the scene classification and recognition model is used to recognize the images collected by multiple cameras, so as to output the road rescue work node, and the road rescue work node includes leaving rescue Stationary node of the workstation, node of arriving at the high-speed rescue site, node of starting rescue trailer work, node of ending the rescue work and leaving the rescue site, or returning to the node of the station of the rescue work station, so as to reduce the error of the output result of the road rescue work node, that is, to improve the road rescue work. The accuracy of the node output results.

It should be noted that the structure of the scene classification recognition model is not limited here. Exemplarily, the scene classification recognition model may include an EfficientNet convolutional neural network model. Due to the high speed and precision of the EfficientNet convolutional neural network model, the training speed of the EfficientNet convolutional neural network model is faster, the training time is shortened, and the recognition accuracy of the EfficientNet convolutional neural network model is higher. The accuracy of the determination result of the rescue work node.

Among them, the EfficientNet convolutional neural network model includes a variety of network structures with different parameters, for example: the EfficientNet convolutional neural network model includes eight network structures with different parameters from EfficientNet-B0 to EfficientNet-B7, and the training of network structures with different parameters The speed and recognition accuracy are different.

It should be noted that each step in this embodiment can be applied to the vehicle-mounted terminal on the rescue vehicle, and the vehicle-mounted terminal can be called an algorithm box for the rescue vehicle to use for localized recognition, and the middle part of the algorithm box can be deployed with scene classification and recognition The scene classification of the model is an offline recognizer, and the network side device can be a rescue monitoring platform, and the vehicle terminal and the rescue monitoring platform can communicate through the network.

It should be noted that the above steps of collecting images, obtaining images, and outputting road rescue work nodes can all be realized on the vehicle terminal, that is, it can be realized offline on the vehicle terminal, and the road rescue The working nodes are uploaded to the network-side device, so that there is no need to communicate with the network-side device before outputting the road rescue working node, thereby reducing the consumption of computing resources and the power consumption of each device.

In the embodiment of the present application, before acquiring images collected by multiple cameras, the method further includes:

Acquire multiple image sets from multiple cameras.

Multiple image sets are input into the initial model for iterative training to obtain a scene classification recognition model, wherein the initial model is an EfficientNet-B2 convolutional neural network model. That is, the EfficientNet-B2 convolutional neural network model is adopted in this embodiment, that is to say, the parameter of the EfficientNet-B2 convolutional neural network model is EfficientNet-B2, and the width coefficient of the EfficientNet-B2 convolutional neural network model is 1.1, and the depth coefficient It is 1.2, the input image resolution is 260*260, and the dropout rate is 0.3.

In addition, the inference speed of the above-mentioned EfficientNet-B2 convolutional neural network model can be less than 100 ms per frame, and the dropout rate of 0.3 refers to the possibility of the EfficientNet-B2 convolutional neural network model recognizing the road rescue work node of each frame of image If the probability is greater than 0.3 (ie 30%), it will be saved, and the probability of the recognition result of the road rescue work node of each frame image is less than or equal to 0.3 will not be saved.

It should be noted that the maximum possibility of the recognition result of the road rescue work node of each frame image can be determined as the recognition result of the frame image, for example: one possibility of the road rescue work node of a certain frame image is 0.5, and the other is 0.5. One is 0.7, and the road rescue work node corresponding to 0.5 can be the station node leaving the rescue work station, and the road rescue work node corresponding to 0.7 can be the node arriving at the high-speed rescue site, then the road rescue work node of the frame image is the arriving high-speed rescue work node field node.

It should be noted that the recognition result of the image collected by each camera needs to correspond to its installation location. If the recognition result obviously does not correspond to its installation location, the recognition result is unreliable. For example: the third image collected by the third camera is that the trailer structure is in In the second position, since the third camera is located on the vehicle body, and the structure of the trailer can only be collected by the second camera located on the chassis, it can be judged that the above recognition result is not reliable at this time.

In the embodiment of the present application, since the initial model is the EfficientNet-B2 convolutional neural network model, and the network model converges faster during the training process, the training speed is improved and the time spent on training is reduced. At the same time, since multiple image sets are used to train the initial model, the images included in each image set can all belong to the same road rescue work node, so that the training speed of the model can be further improved.

It should be noted that the initial model may also adopt a convolutional neural network model with other parameters.

As an optional implementation manner, multiple image sets are collected through multiple cameras, including:

Multiple capture images are acquired by multiple cameras.

A plurality of captured images are classified into a plurality of image sets by an image classification technique.

Among them, the state of the parts of the rescue vehicle in each image set can belong to the same state, and since the image classification effect of the EfficientNet-B2 convolutional neural network model is better, it can be considered that the EfficientNet-B2 convolutional neural network The model has image classification technology, and the initial model uses the EfficientNet-B2 convolutional neural network model.

In this way, the EfficientNet-B2 convolutional neural network model can directly classify images to obtain multiple image sets, without setting up a separate model to classify captured images, which further improves the training speed of the model and reduces the cost of computing resources. At the same time, image classification technology performs feature operations on the entire image, which can greatly reduce the workload of the model in the sample processing stage.

Of course, other models may also be used to classify the captured images, and the model may include image classification technology.

In the embodiment of the present application, the plurality of cameras include a first camera, a second camera and a third camera, the first camera is located inside the cab of the rescue vehicle, the second camera is located on the chassis of the rescue vehicle, and the third camera is located on the on the body.

In the embodiment of the present application, the image is input into the scene classification recognition model, and the road rescue work node is output, including:

At least one of the first image collected by the first camera, the second image collected by the second camera, and the third image collected by the third camera is input into the scene classification recognition model, and the first image, the first image, and the first image are detected and analyzed. The second image and the third image are output to the road rescue work node. The number of the first image, the second image and the third image is multiple.

Exemplarily, with reference to FIG. 2, the first camera 101 can be located on the inner wall of the cab 104, the second camera 102 can be located in the middle of the chassis 105, or be located at the end of the chassis 105 away from the cab 104 (the third camera 103 can be Located above the cab 104 .

In this embodiment, at least one of the first image captured by the first camera, the second image captured by the second camera, and the third image captured by the third camera may be input into the scene classification and recognition model, so that the scene classification and recognition The model can combine the first image, the second image and the third image collected from different locations to determine the road rescue work node, so that the accuracy of the determination result of the road rescue work node can be improved, and the accuracy of the determination result of the road rescue work node can be reduced. error.

At the same time, since the images captured by the first camera, the second camera and the third camera can fully and accurately reflect the current node of the rescue vehicle, the accuracy of the determination result of the road rescue work node is relatively high.

Of course, the camera can also be installed in other locations, and the scene classification and recognition model can also determine the road rescue work node according to the images collected by the cameras in other locations, which will not be repeated here, but can refer to the relevant principles of the above-mentioned embodiment.

In the embodiment of the present application, the first image, the second image and the third image are detected and analyzed, and the road rescue work node is output, including one of the following:

If the scene classification and recognition model detects that a plurality of first images include part of the first target image, and the vehicle speed acquired in real time is greater than the first threshold, then output the node leaving the rescue station, wherein the display of the first target image The content includes the scene of the first target object; the vehicle speed is the vehicle speed of the rescue vehicle. The first target object scene is a scene image in which there is a driver in the cab of the rescue vehicle.

If the scene classification recognition model detects that multiple first images include the first target image and the second target image, and the vehicle speed obtained in real time is equal to the first threshold, the output reaches the high-speed rescue scene node, wherein the second target The display content of the image does not include the first target object scene, and the acquisition moment corresponding to the first target image is located before the acquisition moment corresponding to the second target image;

If the scene classification recognition model detects at least one of the situation that the third target image and the fourth target image are included in the multiple second images, and at least one of the fifth target image is included in the multiple third images, then the output starts to rescue In the trailer working node, the display content of the third target image includes the trailer structure of the rescue vehicle at the first position, the display content of the fourth target image includes the trailer structure at the second position, and the display content of the fifth target image includes the vehicle to be rescued. and at least one of the second target objects; the second target objects may include rescue workers.

If the scene classification recognition model detects that multiple second images include the third target image, the fourth target image and the sixth target image, then output the end rescue work and leave the rescue scene node, wherein the number of the fourth target image greater than the number threshold, the display content of the sixth target image includes the rescued vehicle being located on the rescue vehicle; wherein, when the number of the fourth target image is greater than the data threshold, it indicates that the duration of the trailer structure of the rescue vehicle at the second position is greater than the first duration .

If the scene classification recognition model detects that the seventh target image and the eighth target image are included in multiple first images, the ninth target image is included in multiple third images, and the vehicle speed obtained in real time is equal to the first threshold Next, the output returns to the station node of the rescue workstation, wherein the display content of the seventh target image includes the first target object scene, the display content of the eighth target image does not include the first target object scene, and the acquisition time corresponding to the seventh target image Before the acquisition time corresponding to the eighth target image, the displayed content of the ninth target image includes that the rescue vehicle is not provided with a rescued vehicle.

Wherein, the first target image indicates that there is a driver inside the cab, the value of the first threshold can be 0km/h, the second target image indicates that there is no driver in the cab, and the third target image indicates that the trailer structure is in the initial position (i.e. The first position represents the initial position), the fourth target image represents that the trailer structure leaves the initial position (that is, the second position is a different position from the first position, indicating that the trailer structure is performing a trailer operation), and the fifth target image includes the rescued vehicle and Rescue at least one of the workers, and the trailer structure is located at the second position for longer than the first duration, it can be accurately determined that the trailer structure is indeed performing towing work at this time.

In this embodiment, since different road rescue work nodes have different judgment standards, the road rescue work nodes can be accurately and quickly determined according to the corresponding standards, which improves the accuracy and speed of the determination result of the road rescue work nodes.

In the embodiment of the present application, the method further includes: sending the road rescue working node to the network side device. In this way, the network side device can determine the road rescue work node where each rescue vehicle is located, so as to facilitate the deployment of rescue vehicles, so as to improve the management of rescue vehicles and the deployment efficiency of rescue vehicles.

It should be noted that, while determining the road rescue work node, the time node corresponding to the above road rescue work node can also be recorded, and the road rescue work node and the corresponding time node can be uploaded to the network side device at the same time. Thereby enhancing the monitoring effect and verification efficiency of each rescue vehicle.

Hereinafter, an example is used to illustrate the above-mentioned implementation manner.

The judgment conditions for leaving the station node of the rescue work station are: the vehicle starts to ignite and power on and the vehicle speed is greater than 0km/h (ie, the first threshold), and the camera inside the vehicle (ie, the first camera) detects "inside-people" for multiple consecutive frames Information (that is, the information of the first target image), giving a judgment conclusion that the rescue vehicle has left the station of the rescue work station (that is, at the node leaving the station of the rescue work station) and recording the time node.

The judgment condition for arriving at the high-speed rescue site node is: the vehicle speed is still at 0km/h, and the picture information of the camera in the vehicle changes from the detection result of "inside-person" (that is, the information of the first target image) to "inside-unmanned" (that is, the information of the second target image), give a judgment conclusion that the rescue vehicle has reached the high-speed rescue site node and record the time node.

The judgment condition for starting the rescue trailer work node is: the bottom camera (ie, the second camera) detects "bottom-start operation" (ie, the information of the fourth target image), and when the actual rescue starts, the picture acquired by the bottom camera is usually The trailer board or trailer arm (ie the trailer structure) of the rescue vehicle is moving (that is, including the third target image and the fourth target image, and determining whether the trailer structure is moving according to the third target image and the fourth target image); or The right camera (that is, the third camera) detects "right-start operation". During the actual rescue work, the picture acquired by the right camera usually includes the rescued vehicle parked behind and the rescue workers wearing work uniforms ( That is, the information of the fifth target image). The continuous multi-frame stable detection of these two label categories will be used to determine the start of the rescue trailer work node and record the time node.

The judgment condition for ending the rescue work and leaving the rescue site node is: the bottom camera detects the information of "bottom-working" and "bottom-ending" successively. During the actual rescue work, the working picture acquired by the bottom camera is usually The trailer board or trailer arm of the rescue vehicle has left the original and fixed scene picture (that is, it can be determined whether it is in the above picture by combining the third target image and the fourth target image), and the end operation picture captured by the bottom camera The rescue vehicle places a fixed scene picture (ie, the display content of the sixth target image) on the rescue vehicle. Through the stable detection of these two types of label screens, it will determine the end of the rescue work and leave the rescue site node and record the time node.

The judgment conditions for returning to the stagnant node of the rescue workstation are: the recognition of the front 4 working nodes, the speed of the rescue vehicle is reduced to a standstill, and the offline recognizer of the artificial intelligence (AI) box detects information from the camera inside the vehicle from the "internal - Someone" (that is, the information of the seventh target image) becomes "inside - unmanned" (that is, the information of the eighth target image), and the right camera detects the "right - empty board" picture (that is, the first Nine target image information). Give the judgment conclusion that the rescue vehicle leaves and returns to the station node of the rescue workstation and records the time node.

It should be noted that the image set can refer to the following expressions:

The images of the rescue work images collected by the camera at the bottom of the rescue vehicle (that is, the second camera) are classified into: "bottom-start operation", "bottom-working", "bottom-end operation", "bottom-vehicle suspension". The images of the rescue work images collected by the camera on the right side of the rescue vehicle (that is, the third camera) are classified into: "right side - empty board", "right side - start operation", "right side - end operation". The images of the rescue work images collected by the internal camera of the rescue vehicle (ie, the first camera) are classified into: "inside-people", "inside-unmanned". Finally, a total of 9 categories of image sets were sorted out. The initial model is trained through the above image set, so that the final trained scene classification recognition model can quickly and accurately determine the road rescue work node where the rescue vehicle is currently located by identifying the state of each component of the rescue vehicle.

Referring to FIG. 3 , FIG. 3 is one of the structural diagrams of a device for determining a road rescue work node provided by an embodiment of the present application. As shown in Figure 3, the road rescue work node determining device 200 includes:

The obtaining module 201 is configured to obtain images collected by multiple cameras, wherein the multiple cameras

different positions of the head on the rescue vehicle;

The output module 202 is configured to input the image into the scene classification and recognition model, and output the road rescue work node, wherein the scene classification and recognition model is trained through a plurality of pre-acquired image sets.

In the embodiment of the present application, the plurality of cameras include a first camera, a second camera and a third camera, the first camera is located inside the cab of the rescue vehicle, the second camera is located on the chassis of the rescue vehicle, and the third camera is located on the on the body;

The output module 202 is configured to input at least one of the first image captured by the first camera, the second image captured by the second camera, and the third image captured by the third camera into the scene classification recognition model, and detect and analyze The first image, the second image, and the third image output the road rescue work node, and the number of the first image, the second image, and the third image is multiple.

Wherein, the road rescue work node includes: a node leaving the rescue work station station, a node arriving at the high-speed rescue site, a node starting the rescue trailer work, a node leaving the rescue site after the rescue work, or a node returning to the station station of the rescue work station.

In the embodiment of the present application, the output module 202 is configured to output Leaving the station node of the rescue workstation, wherein the display content of the first target image includes the scene of the first target object;

In the embodiment of the present application, the output module 202 is configured such that if the scene classification recognition model detects that multiple first images include the first target image and the second target image, and the vehicle speed obtained in real time is equal to the first threshold, Then the output reaches the high-speed rescue scene node, wherein the display content of the second target image does not include the first target object scene, and the acquisition time corresponding to the first target image is located before the acquisition time corresponding to the second target image;

In the embodiment of the present application, the output module 202 is configured such that if the scene classification recognition model detects that multiple second images include the third target image and the fourth target image, and that multiple third images include the fifth target In at least one of the situations of the image, output the starting rescue trailer work node, wherein, the display content of the third target image includes that the trailer structure of the rescue vehicle is located at the first position, and the display content of the fourth target image includes that the trailer structure is located at the second position. position, the display content of the fifth target image includes at least one of the rescued vehicle and the second target;

In the embodiment of the present application, the output module 202 is configured to output the end of the rescue work if the scene classification recognition model detects that the third target image, the fourth target image and the sixth target image are included in the multiple second images. Leaving the rescue scene node, wherein the number of the fourth target image is greater than the number threshold, and the display content of the sixth target image includes that the rescued vehicle is located on the rescue vehicle;

In the embodiment of the present application, the output module 202 is configured such that if the scene classification recognition model detects that the seventh target image and the eighth target image are included in the plurality of first images, the ninth target image is included in the plurality of third images, and When the vehicle speed obtained in real time is equal to the first threshold, the output is returned to the station node of the rescue station, wherein the display content of the seventh target image includes the first target object scene, and the display content of the eighth target image does not include the first target object In the scenario, the stimulation time corresponding to the seventh target image is before the acquisition time corresponding to the eighth target image, and the display content of the ninth target image includes that there is no vehicle to be rescued on the rescue vehicle.

In the embodiment of the present application, the scene classification recognition model includes: an EfficientNet convolutional neural network model.

In the embodiment of the present application, the road rescue work node determination device 200 also includes:

A collection module configured to collect multiple image sets through multiple cameras;

The training module is configured to input multiple image sets into the initial model for iterative training to obtain a scene classification and recognition model, wherein the initial model is an EfficientNet-B2 convolutional neural network model.

In the embodiment of the present application, the acquisition module includes:

The collection sub-module is configured to collect multiple captured images through multiple cameras;

The classification sub-module is configured to classify the multiple captured images into multiple image sets through image classification technology.

The sending module is configured to send the road rescue work node to the network side device.

The road rescue work node determination device 200 can realize the various processes of the method embodiment in FIG. 1 in the embodiment of the present application, and achieve the same beneficial effect. In order to avoid repetition, details are not repeated here.

The embodiment of the present application also provides an electronic device. Referring to FIG. 4 , the electronic device may include a processor 301 , a memory 302 and a program 3021 stored in the memory 302 and executable on the processor 301 . When the program 3021 is executed by the processor 301, any step in the method embodiment corresponding to FIG. 1 can be implemented and the same beneficial effect can be achieved, so details are not repeated here.

Those skilled in the art can understand that all or part of the steps for implementing the methods of the above embodiments can be completed by program instructions related hardware, and the program can be stored in a readable medium. The embodiment of the present application also provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, any step in the above method embodiment corresponding to Figure 1 can be realized, and the same Technical effects, in order to avoid repetition, will not be repeated here.

The storage medium is, for example, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and the like.

The above is the preferred implementation of the embodiment of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the application, some improvements and modifications can also be made, and these improvements and modifications should also be considered For the scope of protection of this application.

Industrial Applicability

In the embodiment of the present application, the images collected by multiple cameras are obtained, wherein the multiple cameras are located at different positions on the rescue vehicle; the images are input into the scene classification recognition model, and the road rescue work node is output, wherein , the scene classification recognition model is obtained by training a plurality of pre-acquired image sets. In this way, because the scene classification model is trained through pre-acquired multiple image sets, the multiple image sets include various images collected by the camera on the rescue vehicle, so the output result of the scene classification recognition model is more accurate, which can reduce road The error of the output result of the rescue work node is to improve the accuracy of the output result of the road rescue work node.

Claims

A method for determining road rescue work nodes, comprising:

acquiring images captured by a plurality of cameras, wherein the plurality of cameras are located at different positions on the rescue vehicle;

The image is input into the scene classification and recognition model, and the road rescue work node is output, wherein the scene classification and recognition model is obtained by training a plurality of pre-acquired image sets.
The method according to claim 1, wherein the plurality of cameras comprises a first camera, a second camera and a third camera, the first camera is located inside the cab of the rescue vehicle, and the second camera is located On the chassis of the rescue vehicle, the third camera is located on the body of the rescue vehicle;

The described image is input into the scene classification recognition model, and the output road rescue work node includes:

input at least one of the first image captured by the first camera, the second image captured by the second camera, and the third image captured by the third camera into the scene classification recognition model, and detect Analyzing the first image, the second image and the third image, outputting the road rescue work node, the number of the first image, the second image and the third image is multiple ;

Wherein, the road rescue work node includes: a node leaving the rescue work station station, a node arriving at the high-speed rescue site, a node starting the rescue trailer work, a node leaving the rescue site after the rescue work, or a node returning to the station station of the rescue work station.
The method according to claim 2, wherein said detecting and analyzing said first image, said second image and said third image, and outputting said road rescue work node includes one of the following: if said When the scene classification recognition model detects that some of the first target images are included in the plurality of first images, and the vehicle speed acquired in real time is greater than the first threshold, then output the station node for leaving the rescue work station, wherein the The display content of the first target image includes the first target object scene;

If the scene classification recognition model detects that multiple first images include the first target image and the second target image, and the vehicle speed obtained in real time is equal to the first threshold, then output the arrival The high-speed rescue site node, wherein the display content of the second target image does not include the first target object scene, and the acquisition time corresponding to the first target image is before the acquisition time corresponding to the second target image;

If the scene classification recognition model detects at least one of the situation that the third target image and the fourth target image are included in the plurality of second images, and at least one of the situation that the fifth target image is included in the plurality of third images One, then output the start rescue trailer work node, wherein the display content of the third target image includes that the trailer structure of the rescue vehicle is located at the first position, and the display content of the fourth target image includes the trailer structure Located at the second position, the display content of the fifth target image includes at least one of the rescued vehicle and the second target object;

If the scene classification and recognition model detects that a plurality of the second images include the third target image, the fourth target image and the sixth target image, then output the end of the rescue work and leave the rescue scene A node, wherein the number of the fourth target image is greater than a number threshold, and the display content of the sixth target image includes that the rescued vehicle is located on the rescue vehicle;

If the scene classification recognition model detects that multiple first images include the seventh target image and the eighth target image, multiple third images include the ninth target image, and the vehicle speed obtained in real time is equal to In the case of the first threshold value, then output the return to rescue workstation station node, wherein the display content of the seventh target image includes the first target object scene, and the display content of the eighth target image does not include the In the scene of the first target object, the acquisition time corresponding to the seventh target image is before the acquisition time corresponding to the eighth target image, and the display content of the ninth target image includes that the rescue vehicle is not equipped with the Rescue vehicle.
The method according to claim 1, wherein the scene classification recognition model comprises: an EfficientNet convolutional neural network model.
The method according to claim 4, wherein, before acquiring images collected by a plurality of cameras, the method further comprises:

capturing the plurality of image sets through the plurality of cameras;

The plurality of image sets are input into an initial model for iterative training to obtain the scene classification recognition model, wherein the initial model is an EfficientNet-B2 convolutional neural network model.
The method according to claim 5, wherein said collecting said plurality of image sets through said plurality of cameras comprises:

collecting a plurality of photographed images through the plurality of cameras;

The plurality of captured images are classified into the plurality of image sets by an image classification technique.
The method according to claim 1, wherein the method further comprises:

Send the road rescue working node to the network side device.
A road rescue work node determination device, comprising:

An acquisition module configured to acquire images captured by multiple cameras, wherein the multiple cameras are located at different positions on the rescue vehicle;

The output module is configured to input the image into the scene classification and recognition model, and output the road rescue work node, wherein the scene classification and recognition model is obtained by training a plurality of pre-acquired image sets.
An electronic device, comprising: a transceiver, a memory, a processor, and a program stored on the memory and operable on the processor; the processor is used to read the program in the memory to implement the claim Steps in the method for determining road rescue work nodes described in any one of 1 to 7.
A readable storage medium for storing a program, and when the program is executed by a processor, the steps in the method for determining a road rescue working node according to any one of claims 1 to 7 are realized.